CN108106500B - Missile target type identification method based on multiple sensors - Google Patents
Missile target type identification method based on multiple sensors Download PDFInfo
- Publication number
- CN108106500B CN108106500B CN201711396230.4A CN201711396230A CN108106500B CN 108106500 B CN108106500 B CN 108106500B CN 201711396230 A CN201711396230 A CN 201711396230A CN 108106500 B CN108106500 B CN 108106500B
- Authority
- CN
- China
- Prior art keywords
- matrix
- sample
- vector
- classified
- missile target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F41—WEAPONS
- F41H—ARMOUR; ARMOURED TURRETS; ARMOURED OR ARMED VEHICLES; MEANS OF ATTACK OR DEFENCE, e.g. CAMOUFLAGE, IN GENERAL
- F41H11/00—Defence installations; Defence devices
- F41H11/02—Anti-aircraft or anti-guided missile or anti-torpedo defence installations or systems
Abstract
The invention discloses a missile target type identification method based on multiple sensors, which comprises the following steps: 1) the method comprises the steps that a sensor is used for collecting missile target data, and then a training sample matrix and a test sample matrix are constructed according to the missile target data collected by a plurality of sensors; 2) performing singular value decomposition on the training sample matrix, and determining dimensionality reduction according to the contribution rate of singular values; 3) generating a classification model according to a sample matrix obtained by singular value decomposition, wherein the classification model consists of a linear transformation matrix and a classification center vector group; 4) carrying out projection preprocessing on a sample to be classified to obtain a sample vector to be classified, calculating the cosine distance between the vector and each component in the classification center vector group obtained in the step 3), and taking the group with the closest distance as a type identification result. The missile target type identification method can utilize high-dimensional sample data generated by multiple sensors to carry out missile target type identification with efficiency and accuracy.
Description
Technical Field
The invention relates to a ship combat system technology, in particular to a missile target type identification method based on multiple sensors.
Background
In ship air-to-air self-defense operation, the flight speed, the incidence angle and the terminal maneuvering mode of different types of incoming missiles are different, so different defense operation schemes need to be adopted for different types of missiles. The identification of the type of the target of the incoming missile can improve the pertinence of defense operation.
The requirement of the ship air-to-air self-defense system on the accuracy rate of missile target type identification is increasing. The traditional method for identifying by using single information source assistance cannot meet the requirement of combat use, and the accurate identification of the missile target type by using multiple sensors to perform multi-source information fusion becomes a current research hotspot. However, the data generated by the multiple sensors has high dimensionality, and the data not only contains missile target characteristic information, but also is mixed with redundant and irrelevant information. In a high-dimensional space, the differentiable degree of the distance between different samples is reduced along with the increase of the dimension of sample data, the high-dimensional sample data not only brings exponential increase of the classification calculation complexity, but also reduces the accuracy of a classification algorithm through redundant and irrelevant information between the dimension and the dimension. Therefore, the effect of directly utilizing the data generated by the multiple sensors to identify the type of the missile target is poor.
Disclosure of Invention
The invention aims to solve the technical problem of providing a missile target type identification method based on multiple sensors aiming at the defects in the prior art, and the missile target type identification method based on the multiple sensors is used for carrying out missile target type identification considering both efficiency and accuracy by using high-dimensional sample data generated by the multiple sensors.
The technical scheme adopted by the invention for solving the technical problems is as follows: a missile target type identification method based on multiple sensors comprises the following steps:
1) the method comprises the steps that a sensor is used for collecting missile target data, and then a training sample matrix and a test sample matrix are constructed according to the missile target data collected by a plurality of sensors; the training sample matrix and the testing sample matrix have the same composition structure, each row of the matrix represents a missile target characteristic, and each column represents a missile target sample;
2) performing singular value decomposition on the training sample matrix, and determining dimensionality reduction according to the contribution rate of singular values;
3) generating a classification model according to a sample matrix obtained by singular value decomposition, wherein the classification model consists of a linear transformation matrix and a classification center vector group;
4) carrying out projection preprocessing on a sample to be classified to obtain a sample vector to be classified, calculating the cosine distance between the vector and each component in the classification center vector group obtained in the step 3), and taking the group with the closest distance as a type identification result.
According to the scheme, the sensor in the step 1) comprises a radar detection system and an electronic reconnaissance detection system.
According to the scheme, before singular value decomposition is carried out on the training sample matrix in the step 2), the method further comprises a normalization processing step of the training sample matrix, and the method specifically comprises the following steps:
normalizing each characteristic of the original training sample matrix A according to rows to obtain a sample matrix X, wherein each element of the matrix XElement xijkThe calculation formula of (2) is as follows:
wherein, aijkThe method is characterized in that the method is a method for training the training matrix, and comprises the following steps of obtaining sample values of an original training sample matrix, wherein min is the minimum value of the sample values, max is the maximum value of the sample values, and k is the weight of the characteristic and is given by an expert system or the scores of experts in the field.
According to the scheme, the specific steps of performing singular value decomposition on the training sample matrix in the step 2) and determining the dimensionality reduction according to the contribution rate of the singular value are as follows:
2.1) carrying out singular value decomposition on the sample matrix X, and obtaining three matrixes after decomposing the matrix X: u, S and VTThe formula is as follows:
X＝USVT；
the decomposed matrix U is a feature matrix and is used for reflecting the relation among features, namely the correlation among the features, and each row represents one feature; matrix VTThe sample matrix is used for reflecting the relation among samples, namely the correlation among the samples, and each column represents one sample; the singular value matrix S is used for reflecting the importance of the data object;
2.2) determining the retention order d by using a singular value matrix S:
wherein R is the number of non-zero singular values in the S matrix, xijAn element of S
That is, the information amount larger than the t proportion can be described by using d features as required, the matrix U reserves d rows, the matrix S reserves d singular values, and the matrix VTD columns are reserved, obtained
D is the dimensionality after dimensionality reduction;
namely, it is
According to the scheme, the linear transformation matrix is the U after dimension reductiondAnd SdConstituent linear transformation matrices UdSd -1And the system is used for projecting the sample vector to be identified to a low-dimensional space.
According to the scheme, the classification center vector group pi；
Using reduced Vd TMatrix, and obtaining a classification center vector group pi，Vd TFor each column vector in (1)ikI is 1,2, …, c; k is 1,2, …, n; wherein c is the number of classes of samples, and n is the number of samples in each class;
pithe calculation formula is as follows:
according to the scheme, the step 4) is as follows:
4.1) setting the fused sample vector to be classified as y and carrying out normalization processing on the sample to be classified;
4.2) Using the Linear transformation matrix U in the Classification modeldSd -1And carrying out linear transformation on the sample vector y to be classified, and projecting the sample vector y to be classified to a low-dimensional space to obtain a sample vector y' to be classified after linear transformation.
y′＝yUdSd -1
4.3) calculating cosine distances between the sample vectors to be classified after dimensionality reduction and components in the classification center vector group to obtain c values, arranging classification results according to the sizes, wherein the class corresponding to the value with the largest result is a classification identification result, and calculating the cosine distances by adopting the following formula:
wherein, | | · | | represents solving a 2 norm.
The invention has the following beneficial effects: a classification model is constructed by using the idea of matrix singular value decomposition, based on the physical meanings of each matrix after singular value decomposition, the negative influence on classification identification caused by redundancy and irrelevant information in high-dimensional sample data is obviously weakened through data dimension reduction, and the accuracy of missile target type identification is improved.
Drawings
The invention will be further described with reference to the accompanying drawings and examples, in which:
FIG. 1 is a flow chart of a method of an embodiment of the present invention;
FIG. 2 is a schematic illustration of a sample missile type classification according to an embodiment of the invention;
FIG. 3 is a sample matrix composition diagram of an embodiment of the invention;
fig. 4 is a schematic diagram of feature extraction in the dimension reduction process according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
As shown in fig. 1, the technical scheme of the invention comprises the following three steps:
firstly, training sample data and test sample data are generated according to missile target data acquired by a plurality of sensors.
Taking an embodiment of the method of the present invention as an example, first, 11 types of flight target data are collected according to public data. Then, the missile is divided into 2 major classes according to the flight speed of the missile: subsonic and supersonic missiles, and each of the major classes is further subdivided, and the specific classification is shown in fig. 2.
The following 6 missiles were classified as subsonic missiles. American "whale-trapping forks"; AGM-109 "battle axe" (TALCM); russian "white snake-E"; gull in europe; japanese ASM-1; russian Kh-31P "krypton".
The following 5 missiles were classified as supersonic missiles. 12SSM in Japan; AGM-88 "ham" (HARM) in USA; russian Kh-55A/B; russian Kh-101; india "braumos".
The sensors are radar detection systems and electronic reconnaissance detection systems. The data generated by the sensors includes radar detection target data and electronic reconnaissance target data.
The target characteristics that can be obtained from the radar are: target altitude (km), airspeed (m/s), azimuth, azimuthal velocity, altitude angular velocity, target heading, and the like.
Features that can be obtained from an electronic reconnaissance detection system are: radiation frequency (MHz), signal type, signal carrier frequency, signal pulse width, signal repetition frequency, signal mechanism, etc.
Electronic scout parameter encoding rules: 4 signal types such as pulse wave, continuous wave, interference wave and interference clutter are represented by numbers 1,2,3 and 4 respectively; the numbers 1,2 and 3 respectively represent 3 signal carrier frequencies such as single frequency, diversity frequency, frequency agility and the like; the pulse widths of 2 signals such as single and variable signals are respectively represented by numbers 1 and 2; respectively representing 4 signal repetition frequencies such as single, random hopping according to groups, PD system repetition frequency, PD system high repetition frequency and the like by numbers 1,2,3 and 4; the numbers 1,2 and 3 represent 3 signal structures such as secondary radar response, command signal and intra-pulse modulation.
Taking subsonic missiles "whale-arresting forks" as an example, a part of samples generated according to the 3 σ rule centering on the arithmetic mean of the upper and lower bounds of each feature are shown in table 1. Taking the first sample as an example, some of the characteristics are as follows: target height 22.47; the flying speed is 75.3; the signal type of the electronic investigation is type 2 (continuous wave).
Table 1 sample examples
| Feature(s) | |
Sample 2 | Sample 3 | Sample 4 | Sample 5 |
| Target height | 22.47 | 32.66 | 36.29 | 37.40 | 35.31 |
| Flying speed | 75.30 | 85.74 | 71.43 | 68.83 | 97.43 |
| Azimuth angle | 169.25 | 81.37 | 168.09 | 200.19 | 152.57 |
| Azimuthal velocity | 4.65 | 4.68 | 4.73 | 4.58 | 4.47 |
| High and low angle | 140.38 | 166.59 | 124.80 | 84.93 | 110.81 |
| High and low angular velocity | 5.92 | 5.82 | 5.80 | 5.70 | 5.67 |
| Target course | 44.17 | 23.76 | 146.63 | 103.77 | 135.56 |
| Frequency of radiation | 5.97 | 7.61 | 7.14 | 10.46 | 9.20 |
| Kind of signal | 2 | 2 | 2 | 2 | 2 |
| Signal carrier frequency | 3 | 2 | 3 | 2 | 3 |
| Pulse width of signal | 2 | 1 | 2 | 1 | 2 |
| Repetition frequency | 2 | 3 | 2 | 2 | 3 |
| Signal structure | 2 | 3 | 3 | 3 | 2 |
And secondly, training a classification model according to the sample data.
The composition structure of the sample matrix is specifically as follows: each row of the matrix is a characteristic of one of the flying objects and each column is a sample. 1 characteristic and c categories are provided; for each class of targets, n samples are taken to form a sample matrix, as shown in fig. 3.
By taking subsonic missiles as an example, 6 classes of missiles are listed, and c is 6; each missile generates 200 samples, and n is 200; each sample contains 13 features, l 13. The sample matrix is 13 rows and 1200 columns.
And then carrying out normalization processing on the composed sample matrix.Normalizing each characteristic of the matrix A according to rows to obtain a sample matrix X, wherein each element X of the matrix X isijkThe calculation formula of (2) is as follows:
wherein, aijkThe method is characterized in that the method comprises the following steps of (1) taking sample values, min is the minimum value in the sample values, max is the maximum value in the sample values, k is the weight of the characteristic, the weight is given by an expert system or the scores of experts in related fields, and the scoring principle is as follows: the sensor cannot detect the feature (k ═ 0); the characteristics are basically independent of missile target identification (k)<0.01); the characteristics are weakly related to missile target identification (k is more than or equal to 0.01)<1) (ii) a The characteristics are strongly related to missile target identification (k is more than or equal to 1)<50) (ii) a The characteristic has a decisive influence on missile target identification (k is more than or equal to 50).
After normalization a new sample matrix X is composed.
The sample matrix X is subjected to singular value decomposition. After the matrix X is decomposed, three matrices are obtained: u, S and VTThe formula is as follows:
X＝USVT
the decomposed matrix U is called a feature matrix, which reflects the relationship (rows) between features, i.e., the correlation between features, each row representing a feature; scale matrix VTA sample matrix reflects the relationship between samples (columns), i.e. the correlation between samples, each column representing one sample. The singular value matrix S reflects the importance of the data object.
Dimension reduction is to discard unimportant feature information to filter out irrelevant redundant information and details. Data is mapped from a high dimensional space to a low dimensional space. Each feature in the low-dimensional space is a linear combination of features in the high-dimensional space, the feature after dimensionality reduction is related to each original feature, and a schematic diagram of feature extraction is shown in fig. 4.
And determining the reserved order d by using a singular value matrix S, wherein the size of each singular value in the singular value matrix S represents the contribution degree of each singular value, and the larger the contribution degree is, the larger the value is. R is the rank of the matrix A, namely the S matrixNumber of medium non-zero singular values, xijIs an element of S. Determining the retention order, wherein the specific formula is as follows:
the meaning of the above formula is: with d features, an amount of information larger than t can be described. Matrix U retains d rows, matrix S retains d singular values, matrix VTD columns are reserved, obtained
For the matrix after dimensionality reduction, most of the information of the original features is retained.
Using reduced UdAnd SdForm a linear transformation matrix UdSd -1And the system is used for projecting the sample vector to be identified to a low-dimensional space. Using reduced Vd TMatrix, and obtaining a classification center vector group pi，Vd TFor each column vector in (1)ikI is 1,2, …, c; k is 1,2, …, n. p is a radical ofiThe calculation formula is as follows:
a total of c classification center vectors are obtained. Similarly, as described above for subsonic missiles, 200 columns of samples generated by each class of missiles form a classification center vector group, and the arithmetic mean of the classification center vectors is obtained.
Linear transformation matrix UdSd -1And a group p of classification center vectorsiThe classification models are commonly composed.
And thirdly, classifying and identifying the fused target data to be classified.
Setting the fused sample vector to be classified as y, and the identification process is as follows:
firstly, normalization processing is carried out on a sample to be classified.
Then utilizing a linear transformation matrix U in the classification modeldSd -1And carrying out linear transformation on the sample vector y to be classified, and projecting the sample vector y to be classified to a low-dimensional space to obtain a sample vector y' to be classified after linear transformation.
y′＝yUdSd -1
And finally, calculating cosine distances between the sample vectors to be classified after the dimension reduction and all components in the classification center vector group to obtain c values, and arranging classification results according to the sizes, wherein the maximum result is the classification recognition result.
Wherein, | | · | | represents solving a 2 norm.
It will be understood that modifications and variations can be made by persons skilled in the art in light of the above teachings and all such modifications and variations are intended to be included within the scope of the invention as defined in the appended claims.
Claims (4)
1. A missile target type identification method based on multiple sensors is characterized by comprising the following steps:
1) the method comprises the steps that a sensor is used for collecting missile target data, and then a training sample matrix and a test sample matrix are constructed according to the missile target data collected by a plurality of sensors; the training sample matrix and the testing sample matrix have the same composition structure, each row of the matrix represents a missile target characteristic, and each column represents a missile target sample;
2) performing singular value decomposition on the training sample matrix, and determining dimensionality reduction according to the contribution rate of singular values;
the specific steps of performing singular value decomposition on the training sample matrix in the step 2) and determining the dimensionality reduction according to the contribution rate of the singular value are as follows:
2.1) carrying out singular value decomposition on the sample matrix X, and obtaining three after decomposing the matrix XThe matrix is as follows: u, S and VTThe formula is as follows:
X＝USVT；
the decomposed matrix U is a feature matrix and is used for reflecting the relation among features, namely the correlation among the features, and each row represents one feature; matrix VTThe sample matrix is used for reflecting the relation among samples, namely the correlation among the samples, and each column represents one sample; the singular value matrix S is used for reflecting the importance of the data object;
2.2) determining the retention order d by using a singular value matrix S:
wherein R is the number of non-zero singular values in the S matrix, xiiAn element that is S;
that is, the information amount larger than the t proportion can be described by using d features as required, the matrix U reserves d rows, the matrix S reserves d singular values, and the matrix VTD columns are reserved, obtained
D is the dimensionality after dimensionality reduction;
namely, it is
3) Generating a classification model according to a sample matrix obtained by singular value decomposition, wherein the classification model is composed of a linear transformation matrix and a classification center vector group;
the linear transformation matrix is the U after dimension reductiondAnd SdConstituent linear transformation matrices UdSd -1The system comprises a low-dimensional space and a sample vector to be identified, wherein the low-dimensional space is used for projecting the sample vector to be identified to the low-dimensional space;
the set of classification center vectors piUsing reduced Vd TMatrix to obtain a classification center vector group pi，Vd TFor each column vector in (1)ikI is 1,2, …, c; k is 1,2, …, n; wherein c is the number of classes of samples, and n is the number of samples in each class;
pithe calculation formula is as follows:
4) carrying out projection preprocessing on a sample to be classified to obtain a sample vector to be classified, calculating the cosine distance between the vector and each component in the classification center vector group obtained in the step 3), and taking the group with the closest distance as a type identification result.
2. The method for identifying the type of the missile target based on the multiple sensors according to claim 1, wherein the sensors in the step 1) comprise a radar detection system and an electronic reconnaissance detection system.
3. The missile target type identification method based on multiple sensors as claimed in claim 1, wherein before the singular value decomposition of the training sample matrix in the step 2), the method further comprises a normalization processing step of the training sample matrix, and the normalization processing step specifically comprises the following steps:
normalizing each characteristic according to rows of the original training sample matrix A to obtain a sample matrix X, wherein each element X of the matrix XijkThe calculation formula of (2) is as follows:
wherein, aijkIs the sample value of the original training sample matrix, min is the minimum value of the sample values, max is the maximum value of the sample values, k0The weight of the feature is given by the expert system score.
4. The missile target type identification method based on multiple sensors as claimed in claim 1, wherein the step 4) is as follows:
4.1) setting the fused sample vector to be classified as y and carrying out normalization processing on the sample to be classified;
4.2) Using the Linear transformation matrix U in the Classification modeldSd -1Carrying out linear transformation on the sample vector y to be classified, and projecting the sample vector y to be classified to a low-dimensional space to obtain a sample vector y' to be classified after linear transformation;
y′＝yUdSd -1；
4.3) calculating cosine distances between the sample vectors to be classified after dimensionality reduction and components in the classification center vector group to obtain c values, arranging classification results according to the sizes, wherein the class corresponding to the value with the largest result is a classification identification result, and calculating the cosine distances by adopting the following formula:
wherein, | | · | | represents solving a 2 norm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711396230.4A CN108106500B (en) | 2017-12-21 | 2017-12-21 | Missile target type identification method based on multiple sensors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711396230.4A CN108106500B (en) | 2017-12-21 | 2017-12-21 | Missile target type identification method based on multiple sensors |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108106500A CN108106500A (en) | 2018-06-01 |
| CN108106500B true CN108106500B (en) | 2020-01-14 |
Family
ID=62212123
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711396230.4A Active CN108106500B (en) | 2017-12-21 | 2017-12-21 | Missile target type identification method based on multiple sensors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108106500B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109241231A (en) * | 2018-09-07 | 2019-01-18 | 武汉中海庭数据技术有限公司 | The accurately pretreatment unit and method of diagram data |
| CN113364315B (en) * | 2021-05-08 | 2022-04-12 | 南昌大学 | Three-phase inverter control method based on gull optimization algorithm |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103077402A (en) * | 2012-12-28 | 2013-05-01 | 湖北省电力公司电力科学研究院 | Transformer partial-discharging mode recognition method based on singular value decomposition algorithm |
| CN105005783A (en) * | 2015-05-18 | 2015-10-28 | 电子科技大学 | Method of extracting classification information from high dimensional asymmetric data |
| CN105956611A (en) * | 2016-04-25 | 2016-09-21 | 西安电子科技大学 | SAR image target identification method based on authentication non-linear dictionary learning |
| CN107132516A (en) * | 2017-05-16 | 2017-09-05 | 电子科技大学 | A kind of Radar range profile's target identification method based on depth confidence network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8879793B2 (en) * | 2013-02-20 | 2014-11-04 | Raytheon Company | Synthetic aperture radar map aperture annealing and interpolation |
-
2017
- 2017-12-21 CN CN201711396230.4A patent/CN108106500B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103077402A (en) * | 2012-12-28 | 2013-05-01 | 湖北省电力公司电力科学研究院 | Transformer partial-discharging mode recognition method based on singular value decomposition algorithm |
| CN105005783A (en) * | 2015-05-18 | 2015-10-28 | 电子科技大学 | Method of extracting classification information from high dimensional asymmetric data |
| CN105956611A (en) * | 2016-04-25 | 2016-09-21 | 西安电子科技大学 | SAR image target identification method based on authentication non-linear dictionary learning |
| CN107132516A (en) * | 2017-05-16 | 2017-09-05 | 电子科技大学 | A kind of Radar range profile's target identification method based on depth confidence network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108106500A (en) | 2018-06-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107728142B (en) | Radar high-resolution range profile target identification method based on two-dimensional convolutional network | |
| CN107977642B (en) | High-resolution range profile target identification method based on kernel self-adaptive mean discrimination analysis | |
| CN110018453B (en) | Intelligent model identification method based on aircraft track characteristics | |
| CN109977871B (en) | Satellite target identification method based on broadband radar data and GRU neural network | |
| CN103226196B (en) | Radar target recognition method based on sparse feature | |
| CN107728143A (en) | Radar High Range Resolution target identification method based on one-dimensional convolutional neural networks | |
| CN104732224B (en) | SAR target identification methods based on two-dimentional Zelnick moment characteristics rarefaction representation | |
| CN110109110B (en) | HRRP target identification method based on priori optimal variation self-encoder | |
| CN110907908B (en) | Navigation radar signal sorting method based on envelope analysis | |
| CN111401168B (en) | Multilayer radar feature extraction and selection method for unmanned aerial vehicle | |
| CN103954935A (en) | Radar signal sorting method under minimum L1 bound norm | |
| CN108106500B (en) | Missile target type identification method based on multiple sensors | |
| CN111175718A (en) | Time-frequency domain combined ground radar automatic target identification method and system | |
| Sehgal et al. | Automatic target recognition using recurrent neural networks | |
| CN111368653B (en) | Low-altitude small target detection method based on R-D graph and deep neural network | |
| CN105373809B (en) | SAR target identification methods based on non-negative least square rarefaction representation | |
| CN116304966A (en) | Track association method based on multi-source data fusion | |
| Yuan et al. | Robust attack on deep learning based radar hrrp target recognition | |
| CN114137518A (en) | Radar high-resolution range profile open set identification method and device | |
| CN110880012B (en) | Inter-pulse agile radar radiation source frequency information correlation method for multi-reconnaissance platform | |
| He et al. | Warship formation extraction and recognition based on density‐based spatial clustering of applications with noise and improved convolutional neural network | |
| CN114359743B (en) | Low-slow small target identification method and system based on multiband | |
| CN114492505B (en) | Air group target and extension target identification method based on semi-actual measurement data | |
| CN109932714B (en) | High-resolution echo recovery method for morphing aircraft based on convolutional neural network | |
| CN114065810B (en) | Unmanned aerial vehicle and armed individual soldier classification and identification method based on number of spectral peaks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |